Apple M5 Ultra for local LLMs

9 measured results from 3 builds, each linked to its source.

Can it run? Mistral Medium 3.5 128B on 2x Apple M5 Ultra Mistral Medium 3.5 128B on 4x Apple M5 Ultra Mistral Medium 3.5 128B on Apple M5 Ultra Muse Spark Open 816B-A42B on 2x Apple M5 Ultra Muse Spark Open 816B-A42B on 4x Apple M5 Ultra Qwen3.8 27B on 2x Apple M5 Ultra Qwen3.8 27B on 4x Apple M5 Ultra Qwen3.8 27B on Apple M5 Ultra

All builds with the Apple M5 Ultra

All measured results

Hide speculative-decoding results ⚡ (MTP, DFlash, DSpark and similar draft-and-verify methods give much higher tok/s on predictable text and are not comparable 1:1 with plain decode)

All Apple M5 Ultra builds: see also all GPUs · best LLMs by memory size · download the dataset.

FAQ

How fast is the Apple M5 Ultra for local LLMs?

The fastest plain-decode result we have is 50.89 tok/s on Qwen3.8 27B (BF16, 8,192-token context, MLX, 4-node tensor parallel) in 4x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open at 36.7 tok/s. Speed depends heavily on the model, quant and engine, so compare like with like in the table. RigBase lists 9 measured results from 3 builds.

How much memory does the Apple M5 Ultra have?

unified memory.

Which models have been measured on the Apple M5 Ultra?

Mistral Medium 3.5 128B, Muse Spark Open 816B-A42B, Qwen3.8 27B.

Where do these numbers come from?

Every row links to a build page with its original public source (benchmark logs, READMEs, forum posts). RigBase does not re-run the benchmarks. Results marked ⚡ use speculative decoding and are not comparable 1:1 with plain decode.